A Hybrid AHP-Entropy and K-Means DRASTIC Framework for Geoelectrical-Based Groundwater Vulnerability Mapping in Akwa Ibom State, Nigeria

Abstract

The assessment of groundwater vulnerability to anthropogenic contamination is critical for water resource management, yet traditional empirical overlay index models like DRASTIC suffer from rigid subjective weightings and arbitrary hazard classifications that fail to capture spatial heterogeneity. To overcome these limitations, this study develops and evaluates an integrated hydrogeophysical framework that combines AHP-entropy weighting with unsupervised K-means clustering within the DRASTIC model. A comprehensive field survey comprising 30 Vertical Electrical Sounding (VES) stations and 16 high-resolution 2D Electrical Resistivity Tomography (ERT) profiles was conducted across a complex structural transition zone within the southeastern extension of the Niger Delta Basin, Nigeria (covering Itu, Ibiono Ibom, Ikono, and Ini Local Government Areas). The geoelectric models were tightly constrained by local deep borehole lithology logs to differentiate protective clay aquitards from highly transmissive, vulnerable sand units. Substantive framework optimization was achieved using a hybrid Multi-Criteria Decision Analysis and Machine Learning (MCDA-ML) approach, multiplying expert-derived Analytic Hierarchy Process (AHP) weights with objective spatial variance coefficients derived from Shannon’s Information Entropy. To establish objective groundwater vulnerability zones, the K-means clustering algorithm was applied to the final continuous DRASTIC vulnerability index scores. Rather than relying on subjective manual thresholding, K-means was utilized to naturally partition the continuous vulnerability index into three distinct categories: low, moderate, and high vulnerability. This clustering was executed on the 30 m × 30 m raster cells encompassing the entire study area. Consequently, the classification provides a continuous spatial vulnerability assessment across the region, rather than being restricted solely to the 30 Vertical Electrical Sounding (VES) station coordinates. This processing was validated using the Elbow Method, Silhouette Coefficient (0.62), Calinski-Harabasz Index (108.03), and Davies-Bouldin Index (0.44). High-vulnerability hotspots were isolated in the southern plain sand corridors where highly porous sand matrices offer negligible contaminant attenuation. Moderate vulnerability constitutes the dominant regional background matrix covering intermediate terrains. Isolated low-vulnerability shields were mapped in the northern sector, geologically sustained by dense, low-permeability clay-rich lenses within the Ameki and Imo Shale Formations. The Impact of the Vadose Zone exhibited the highest mean effective weight in the single-parameter analysis, indicating strong localized influence on the vulnerability index. Depth to Water (with Maximum effective weight of 54.72%), exhibited the highest mean sensitivity in the map-removal analysis, indicating that it was the most influential parameter in the overall index under the adopted weighting scheme. Conversely, Hydraulic Conductivity perfectly stabilized at 0.00% variation across both diagnostic routines (Map Removal Sensitivity Analysis and Single-Parameter Sensitivity Analysis), demonstrating that the hybrid methodology successfully neutralized non-informative spatial parameters. The resulting field-calibrated framework provides a potentially transferable approach for groundwater vulnerability assessment and aquifer protection in heterogeneous sedimentary basins.